Unfortunately I don't believe there is no one benchmark that will show stability 100%. I saw the same thing on the ICFX3200. Could run any stress test program there was with 3dMark looping at the same time for hours on end. But Prime would fail in under 3 hours.

I think the days of Prime stability ended quite a while ago. Not only is it no longer the definitive test of stability any longer but its extremely hard on the board and components to constantly be hammered like that.